On Mon, May 04, 2020 at 11:06:48AM +0100, Peter Maydell wrote: > On Mon, 4 May 2020 at 10:57, Michael S. Tsirkin <[email protected]> wrote: > > > > ./scripts/checkpatch.pl --mailback base.. > > > > 2/7 Checking commit 5554e78b18ea (nvdimm: Use configurable ACPI IO base > > > and size) > > > ERROR: Do not add expected files together with tests, follow instructions > > > in tests/qtest/bios-tables-test.c: both > > > tests/qtest/bios-tables-test-allowed-diff.h and hw/acpi/nvdimm.c found > > > > This beats me. Where did we get > > tests/qtest/bios-tables-test-allowed-diff.h from? > > It's a different patch, isn't it? > > Ah, this is a bug in the checkfilename() function -- it uses > some globals $acpi_testexpected and $acpi_nontestexpected, but > there is no code to reset these when checkpatch starts checking > a new patch. So if you only check one patch in a checkpatch run > (eg by just passing it a single patch file) then it will work, but if > a single checkpatch execution is checking several commits > (eg in the way patchew runs it to check the whole series of > git commits at once, or if you pass it several patch files) then > it will give wrong results for the second and later patches. > I think the variables need to be reset at the top of 'sub process()'. > > thanks > -- PMM
Good point. Will fix.
